441

TAXONOMY AND EVOLUTION IN REDPOLLS

Carduelis flammea-hornemanni;A MULTIVARIATE STUDY OF THEIR BIOMETRY

MARC HERREMANS

ABSTRACT Multivariate analyses were performed on four standard mea­surements (wing-chord, tail-length, bill-length and bill-depth) of over 1000sexed Redpolls, including samples from all currently recognized taxa.Samples of flammea and exilipes, phenotypically defined by colour wereseparable biometrically, as were samples of rostrata and hornemanni. DarkRedpolls from Iceland were almost indistinguishable from rostrata, but theiraverage biometry was different from the pale birds breeding on the sameisland. Furthermore, pale birds from Iceland are different from the allopatricpale taxa exilipes and hornemanni, though they are closer to exilipes. Thepale phenotype from Iceland is the most distinct Redpoll on that island, andmight deserve a new name; the name islandica should be restricted to thedark, rostrata-like birds. Current taxonomic arrangements of Redpolls intwo polytypic species are rather poorly founded, because most similaritiesin morphological criteria may be the result of convergent evolution. Mainlybased on zoogeographical argument, specific rank could be assigned to hor­nemanni, exilipes, rostrata, andflammea.

National Fund for Scientific Research, Belgium, Prinses Lydialaan 65,B-3030 Heverlee.

INTRODUCTION

All Redpolls (Carduelis, subgenus Acanthis, spe­cies group flammea- hornemanni) are very alikeand the radiation of these birds is either recent orvery conservative. The taxonomy of Redpolls pro­vokes much debate: some authors have argued toplace all Redpolls under a single specific name,flammea (e.g. Salomonsen 1951, Troy 1985), buton the other hand, Todd (1963) proposed to splitthem in four separate species: hornemanni, exili­pes, rostrata, andj7ammea. Currently two polyty­pic species,flammea and hornemanni are generallyrecognized (e.g. Voous 1977, Wolters 1979, Knox1988). Molau (1985) found no signs of hybridiza­tion between Cf j7ammea and C.h.exilipes duringan extensive field study in Sweden, and Knox(1988) rejected all claims of hybridization betweenC.f.flammea and C.h.exilipes, and betweenCf.rostrata and C.h.hornemanni in his excellentreview, thus supporting the arrangement of Red­polls in two polytypic species. Although Knox cla-

Received 1 February 1989, accepted 10 June 1989.

rified many confusing situations (exilipes-flam­mea, holboellii, rostrata-hornemanni), he basedhis conclusions mainly on the study of plumages.Only univariate statistics were used to corroboratethe findings also biometrically, which lead in somecases to somewhat weaker biometrical interpreta­tions.

In this paper results of multivariate analyses onfour external standard measurements of Redpollsare reported. The conclusions of Molau (1985) andKnox (1988) on the taxonomy of exilipes vs.j7am­mea, rostrata vs. hornemanni and holboellii arecorroborated and extended. The situation in Ice­land and the radiation and relationships amongRedpoll taxa are discussed.

MATERIALS AND METHODS

SamplingMeasurements of Lesser (cabaret) and Mealy

Redpolls (j1ammea) were sampled during ringing

ARDEA 78: 441-458

442 ARDEA 78 (3), 1990

activities from birds on passage or wintering inHeverlee, central Belgium (500 50' N; 40 40' E)during the years 1976-1986; many Mealy Redpolls,including about 8.5% "holboellii"-forms werecaught during a huge irruption to western Europein late autumn 1986. Individuals of cabaret origina­ted from Britain, Denmark and the low countries.Specimens representing all taxa were also studiedfrom museum skins during visits to the British Mu­seum (Natural History) (BMNH) at Tring, the Zoo­logisk Museum (ZMK) at K0benhavn and the Ko­ninklijk Belgisch Instituut voorNatuurwetenschap­pen (KBIN) at Brussels. Further specimens wereobtained on loan from the American Museum ofNatural History (AMNH), New York, the RoyalMuseum of Scotland (RMS), Edinburgh, the Sta­vanger Museum (SMN), Stavanger, the Zoologi­sches Museum der Humboldt Universitat (ZMHU),Berlin, and the Icelandic Museum of NaturalHistory (IMNH), Reykjavik. For accuracy, onlymeasurements of specimens with well prepared,closed bills were used. No adjustment was made forthe obvious increase in bill-length found during thebreeding season in some specimens from the taxaflammea, exilipes, cabaret, rostrata and islandica.The effect of age, which is generally not significanton the biometry of Redpolls (Knox 1988) was nottaken into account when analyzing biometricaldata. Mixing of measurements from live birds andskins only causes minor influences on the statisticalprocedures generating population estimates fromsamples (see Herremans 1985);it increases intra-group variation, resulting insmaller statistical discrimination between groups.

For the taxa rostrata, islandica, and horneman­ni, only specimens from the breeding areas wereused. According to recent views that immature andfemale exilipes generally are quite streaked on therump and underside (Molau 1985, Knox 1988), birdsapparently misidentified as to taxon were stored ina separate group ("unknowns"; 20 in total), andwere not included inflammea or exilipes.

Specimens from Iceland were separated intothree plumage classes: birds were considered "pale"when they showed pale overall plumage with whi­tish rump and underparts, at most moderately stre-

aked in adults; birds were considered "dark" whenboldly streaked on partly buffish-brown underpartsand boldly streaked on dark, brownish upperpartsbut lacking the white rump; birds showingcombinations of features of both extremes wereclassified as "indeterminate". Measurements werenot used to allocate taxa, though previous workersresponsible for the identification on the labels mighthave done so.

Sexing oflive birds was based on plumage char­acters (see Appendix). Sex of museum skins wastaken from the labels. It is possible that up to 10%were erroneously sexed, but this is of little impor­tance to the results reported below. The numbers ofspecimens used in the analyses are given in Table1: males were predominant.

MeasurementsFour measurements were taken: maximum

wing-length, bill-length to feathering, and bill­depth at feathering were sampled according to pro­cedures of Svensson (1984). Tail-length, however,was not measured with a ruler, but by inserting oneend ofa pair ofpointed calipers between the centralpair of rectrices, a procedure more easily appliedto skins. Live birds were also weighed, but thisparameter was only used in the investigation of"holboellii". The case of"holboellii" has been con­clusively reviewed by Molau (1985) and Knox(1988), but this group is discussed in this paper aswell, because of the implications it has on the treat­ment of flammea as a group.

AnalysesPrincipal component analysis (PCA) was per­

formed on the correlation matrix, resulting in varia­tion within the small variables (bill-length and bill­depth) having a similar level of importance as vari­ation in the larger variables (wing and tail). PCAwas used to draw a general view of all Redpoll taxain the plane of most biometrical variation amongall individuals. The eigenvalue of a principal com­ponent (axis) represents the amount of variationpresent in the direction of the corresponding eigen­vector, itself a calculated linear function of the ori­ginal variables.

TAXONOMY IN REDPOLLS 443

Table 1. Number of Redpoll specimens used in the analyses.

Taxon Males Females Sum

homemanni 47 42 89pale islandica 21 5 26indeterminate islandica 40 23 63dark islandica "breeding" 14 11 25

"not breeding" 22 20 42rostrata "breeding" 58 25 83

"not breeding" 43 45 88exilipes 47 37 84cabaret 66 34 100flammea 222 162 384"holboellii"-form 31 23 54"unknown" (j1arnmea or exilipes) 7 13 20

Total 618 440 1058

Canonical variate (Canovar) analyses wereused to investigate and visualize the maximum bio­metrical distance between some taxa or subsam­pIes identified a priori on colour. It can also illus­trate the position of individuals in the plottingsrepresenting the maximal separation of the taxa.When the total eigenvalue of an analysis is < I, thesamples included in the analysis are accepted to befrom only one statistical population. Critical sta­tistics were evaluated by nonparametric tests onthe scores of a discriminant function analysis. Bythis procedure DFA is only used to transform theoriginal measurements into scores that maximizethe differences between populations. The implica­tions of required multinormality to perform multi­variate statistics are avoided this way (Blackith &Reyment 1971).

PCA and canovar routines were written for PCby Dr. D. Meirte, following methods detailed inGreenstadt (1967), Sokal & Rohlf (1969) and He­brant (1974).

The plot include sample means, centroids (95%probability predictions of the position of the pop­ulation mean) and 95% equiprobable ellipses forthe position of the whole population.

As a rule of thumb, nearly all individuals ofeach population can be identified when populationellipses fall apart (at most, 5% of the individuals

are at risk of mis-identification). When centroidsfall apart, the means are distinct.

RESULTS

A general over-view of the taxaFigure I illustrates the results of a PCA in­

cluding all specimens from Table I. Sexual dimor­phism is well marked and significant in all popu­lations, except rostrata. It is possible that the varia­tion in the occurrence of red in males of rostratahas caused confusion in the past and introducedsystematic errors in the sexing indicated on thelabels. Several males of rostrata lack red on theunderside, as mentioned by Knox (1988), but somefemales do show a considerable amount of red. Asa consequence of the relatively more frequent mis­sexing in this taxon, the variation within the sampleof each sex may have increased and the differencesamong the sexes may have been obscured.

Two main isometrical chains can be recognizedin Fig. I, both roughly parallel to sexual dimor­phism. In parallel to the general duality in plumageappearance among taxa (pale vs. dark; Knox 1988)there is, however, also a biometrical duality (contraKnox 1988): there is a chain of populations to theleft, including exilipes, pale islandica and horne-

444 ARDEA 78 (3),1990

70~ ~---r---....-16 -12 -8

P.C. 2 (K= 1.05)Fig.I. Plot ofprincipal component analysis on the cor­relation matrix of4 measurements (wing, tail, bill-lengthand -depth) of all Redpoll specimens shown in Table 1.PC1 = 60.6 % and PC2 = 26.3 % of total variation. Theright cline is defined by the mean of cabaret andflam­mea and the left by the mean ofexilipes and hornemanni.Means (spots )and centroids indicated. Codes: f = fe­males, m=males; C =cabaret, F =flammea, E= exilipes,R = rostrata, D = dark "islandica", P = pale "islandica",H = hornemanni.

Bill

length depth

0.24 0.880.96 0.10

-0.97 0.520.22 0.84

-0.83 0.980.54 0.18

0.95 0.99-0.30 0.12

Wing Tail

Table 2. Correlations between the original measure­ments and the first two multivariate axes. A: Principalcomponent analysis on all data (Fig. 1). B: Canonicalvariate analysis on males and females exilipes versusflammea (Fig. 2). C: Canonical variate analysis on malesand females rostrata versus hornemanni (Fig. 4). D:Canonical variate analysis on males and females paleversus dark islandica (Fig. 6).

manni, and a right one, including cabaret andflammea; rostrata and dark islandica are in be­tween (Fig. 1). Distinction between the chains ofpopulations comes along the second principal axis,reflecting the allometry of tail-length (coefficientofdetermination 11.9%), and especially bill-length(coefficient of determination 92.1 %). The leftchain contains the relatively short-billed, but long­tailed birds, and the right chain consists of longer­billed, shorter-tailed birds, as can be inferred fromthe correlation between the original variables andthe principal component axes (Table 2A). At firstsight one could recognize in this figure a corrobo­ration of current Redpoll taxonomy: cabaret,flammea and rostrata in the dark species (flam­mea), and exilipes and hornemanni in the pale spe­cies (hornemanni) (as e.g. in Voous 1977, Wolters1979, Knox 1988). However, the finding ofdark is­landica, currently placed inflammea, near the rightchain, and pale birds from the same population inthe left chain is most intriguing and needs a moredetailed analysis. The figure also shows that the

A. PC 1 0.90 0.88PC2 -0.02 -0.35

B. CV 1 -0.47 0.90CV2 0.88 0.43

C. CV 1 0.97 0.99CV2 0.23 -0.10

D. CV 1 0.37 -0.42CV2 0.93 0.90

ern

85

80

75

- 95Nv.NII~-.... 90.00:

TAXONOMY IN REDPOLLS 445

Fig. 2. Plot of canonical variate analysis on wing, tail,bill-length and -depth offlammea (excluding "holhoel­Iii") males (Fm, open triangles, N = 222) and females(Ff,dark triangles, N = 162) versus exilipes males (Em,open circles, N =47) and females (Ef, dark spots, N =37).Sample means (+), centroids and 95% equiprobable pop­ulation ellipses shown. Total eigenvalue: K = 2.46.

above, "holboellii" is not included injlammea inthis analysis. When "holboellii" individuals areplotted in Fig. 2, they are mainly situated at theupper left edge of flammea and are far separatedfrom exilipes: their absence from the analysis doesnot bias the interpretation of overlap between thetaxa (see also Fig.7).

Biometrical differences agree with colour phe­notypes, and I corroborate Molau's and Knox'sview thatflammea and exilipes are a pair of siblingspecies (see also Mayr & Short 1970), for whichcertain measurements are as useful for identifica­tion than most plumage characteristics. Some spe­cimens, however, appear as yet impossible to iden­tify, but multivariate biometrical overlap on thesefour measurements, as shown, is only about 5%.The samples used are composed of specimens fromdifferent seasons, different geographical origin,different preparation, which makes them subject toartificially increased intra-taxon variation, result­ing in reduced statistical discrimination. Carefullysexed samples from one season and one area willprobably be separable at about 1% level.

two pairs of sibling taxa (sensu Mayr 1963, see alsoMayr & Short 1970) have a tendency to be separa­ted by different mechanisms: rostrata and horne­manni mainly differ in the direction of PC I,representing size, but flammea and exilipes aresegregated by PC2, indicating different propor­tions.

The "holboellii" taxonIn my sample of irruptive Mealy Redpolls from

Belgium, several large-billed individuals werefound. The frequency distribution of bill-lengthsshowed an important skewness (1.33 in 232 malesand 1.43 in 171 females), resulting in large coeffi­cients of variation (SD/Mean in males 8.1 %, in fe­males 9%). When "holboellii" males are definedby bill-length:2:10.5mm and females by bill-length:2:lOmm (these limits mainly cut offthe extra-sym­metrical part from the frequency distribution ofthismeasurement), 40 birds (8.6%) out of a total samp­le of 467 flammea were "holboellii".

However, a canovar analysis on the remainingmeasurements (wing, tail, bill-depth and weight)did not reveal any distinction between these sub­samples (eigenvalue K = 0.39; Herremans 1987)."holboellii" only represents an extreme skew oftheflammea population; in the characters examinedthe effect is mainly restricted to bill- length. Forthe statistical analyses"holboellii" normallyshould be included in jlammea. However, theskewness introduced by doing so greatly enhancesthe variation withinflammea, resulting in less sta­tistical discrimination and perhaps in weaker re­sults because the multivariate normality conditionswon't be met anymore.

The taxaflammea and exilipesA canovar analysis (in casu DFA) of only four

standard measurements shows samples ofexilipesand flammea, phenotypically defined on colour,also to be biometrically distinct (eigenvalue K =2.46; Fig. 2). Table 2B shows that the short bill andthe long tail in particular characterize exilipes, asreported by Molau (1985). Sexual dimorphism is amatter of size, especially of wing length and billdepth. For methodological reasons explained

~

'" Fma"

2.1

~

'":>c.5 2.0

1.9

1.8 Ff

o 0.1 0.2

Em

Ef

0.3

C.V.l IK=2.02)

0.4

446 ARDEA 78 (3), 1990

C.V.1 (K=2.02)

2.22.1

HI

2.01.91.81.7

1.8....0II 1.7~

'":> 1.6ti

1.5

1.4

1.3

C.V.1 (K=2.38)

Fig. 4. Plot of canonical variate analysis on wing, tail,bill-length and -depth of rostrata males (Rm, opentringles,N= 101) and females (Rf, dark triangles,N= 70)versus hornemanni males (Hm, open circle, N = 47) andfemales (Hf, dark spots, N = 42). Sample means(+), cent­roids and 95% equiprobable population ellipses shown.Total eigenvalue: K = 2.55.

The situation in IcelandUp to now, only one taxon has been admitted

as breeding in Iceland, islandica, although it hasbeen known for a long time that this includes birdswith plumages ranging from very dark to very pale.On plumage, dark birds breeding in Iceland appearalmost indistinguishable from rostrata (Knox1988). Some pale breeding birds are very white andalmost indistinguishable from the palest horne­manni or exilipes (Knox 1988), but most are con­siderably more streaked, rather like darker individ­uals of exilipes. Many of my "indeterminate" spe­cimens are more greyish on the back than dark birdsor rostrata and they are also slightly less boldlystreaked, looking more likeflammea. It was gener­ally easier to classify birds as dark than to decidebetween pale or indeterminate, even when theeffect of age on plumages could be taken into con­sideration. The biometrical affinities ofdark islan­dica to rostrata andflammea were investigated.

mainly based on overall size and on the short billofhornemanni; the relative difference in tail-lengthis not important to separate these taxa (Table 2C).

G 0

S-oC

0 -1.9

Ff1.8 EI

0 0.1 0.2 0.3 0.4

Fig. 3. Plot of the "unknown" flammea or exilipes spe­cimens in the canovar figure of flammea and exilipes.a: flammea on the label, but re-identified as exilipes;b: labelled or suggested as (possible) hybrid, but re-iden­tified as exilipes; c: labelled or suggested as (possible)hybrid, but re-identified as flammea; d: exilipes on thelabel, but re-identified as probableflammea; underlinedare males, otherwise females.

The taxa rostrata and hornemanniAlthough these two sibling taxa are generally

considered less of a taxonomic problem, the over­lap is considerable (ca. 35%) in a canovar of onlyfour measurements (Fig. 4). The separation is

In Fig. 2 four specimens offlammea are ratherdistant from their own taxon but are among exili­pes. Although the plumage of these birds was asfar as could be judged from the state of the skinwithin the range variation offlammea (the speciesindicated on the label), they were in fact more like­ly exilipes. When the specimens which I conside­red to have the wrong specific identity on the label(based on plumage appearances) are plotted in Fig.2, they also tum out to be mainly exilipes females,although they were originally identified as flam­mea (Fig. 3). This is in agreement with recent viewsthat variability in exilipes has been underestimatedand that (juvenile) females of exilipes in particularcan be quite streaked and very flammea-like(Molau 1985, Knox 1988).

..V

Fm0 EmII 2.1

~

'":>ti 2.0

TAXONOMY IN REDPOLLS 447

To avoid specimens of rostrata (possiblywintering on Iceland) being included in darkislandica, the canovar analysis was performed onsamples which were also split by season: speci­mens collected from May to August were conside­red as "breeding birds" and those from Septemberto April as "wintering birds". Samples of "bree­ding" and "wintering" birds did not significantlydiffer for dark islandica or for rostrata (Fig. 5).Although some other sample means are distinct at5% level (e.g. Dfb vs. Rfb and Dfvs. Rf), canovarshows that all these 8 predicted populations ofdarkspecimens could belong to one statistical popula­tion (Fig. 5, total eigenvalue K = 0.77). It can alsobe seen that differences between origins are smallerthan between sexes. DFA on the males, the mostcritical groups because they contain mainly cor­rectly sexed birds, shows no separation between

-0.7

'"'"oIIlC

the populations predicted from the samples (K =0.21), yet the sample medians of DFA scores areclearly distinct (Mann-Whitney-U test Z =4.8, P< 0.00001). The population of dark islandica is asfar as these 4 standard measurements are concernednot separable from rostrata, but it may be more dis­tinct on criteria not included here. However, darkislandica can clearly be separated fromflammea(eigenvalue K = 2.23, figure not shown).

Separation between populations predicted forpale and dark birds from Iceland is marginal (Fig.6), but sample medians of DFA scores are highlysignificantly different between pale and dark males(Mann-Whitney-U testZ =5.9,P < 0.000001). Theseparation is mainly based on the smaller bill andthe somewhat longer tail of the pale birds (Table2D). Sexual dimorphism is smaller than the differ­ence between the pale and dark forms. When inde­terminate birds are plotted in the canovar of paleand dark forms, they associate with the pale ratherthan with the dark birds (Fig. 6), from which theirmedian scores of separate DFAS are also more dis-

-1.0+-----,-----,------.--

:> -0.8c.5

-0.9

.......;:..I'". Ii."

._ .. ;+ ....\

01 .

2.3

RIb

~.~... ,

2.4 2.5

C.V. 1 (K=0.49)

N 3.5

"ci

"lC'" 3.4:>J

3.3

3.2

Pf

+

•.........•.·······.. Pm

Fig. 6. Plot of canonical variate analysis on wing, tail,bill-length and -depth of dark males (Dm, N = 36) andfemales (Df, N =31) (full line) versus pale males (Pm,N = 21) and females (Pf, N = 5) (dotted) from Iceland.Indeterminate males (1m, N = 40) and females (If, N = 23)afterwards plotted (dashed) following the canovar func­tions of pale versus dark. Only sample means (+) andcentroids shown. Total eigenvalue: K = 1.01.

Fig. 5. Plot of canonical variate analysis on wing, tail,bill-length and -depth of dark islandica versus rostrata:Dfb =dark islandica females breeding season, N = 11;Dmb = dark islandica males breeding season, N = 14;Df = dark islandica females outside breeding season,N = 20; Dm = dark islandica males outside breeding sea­son, N = 22; Rfb = rostrata females breeding season,N = 25; Rmb = rostrata males breeding season, N = 58;Rf = rostrata females outside breeding season, N = 45;Rm = rostrata males outside breeding season, N = 43;Only sample means (+) and centroids shown. Totaleigenvalue: K = 0.77.

3.1

2.3 2.4 2.5 2.6 2.7

C.V. 1 (K=0.59)

448 ARDEA 78 (3), 1990

-o.a.L----.--------.-----,---=-----,.--

DISCUSSION

siderable overlap exists at population level andmany individuals cannot be separated biometrical­ly.

3.0

,, H

\\

\III

I/

//

"

C.V. 1 (K= 5.76)

2.a

\\ ., ".>..'

2.62.4

~ -0.5

Fig. 8. Plot of canonical variate analysis on wing, tail,bill-length and -depth of the three pale taxa: exilipes (E,full line, N =84), pale islandica (P, dotted & dashed, N= 26) and hornemanni (R, dashed, N = 89). Samplemeans (+), centroids and 95% equiprobable ellipsesshown. Total eigenvalue: K = 5.80.

-0.7

-0.6

~ -0.3ooII

~ -0.4

tinct. This agrees with the remark above that it wasmore difficult to assign a specimen as pale or inde­terminate rather than as dark: the indeterminategroup presumably contains many first year palebirds which are more streaked, and some old, paler,genetically dark birds. Although the sample size ofpale females is very low for a multivariate analysis,it is obvious from the neat position of the popula­tion predicted from this sample (Fig. 6) that thisgroup did not distort the analysis.

Furthermore, at population level, pale birdsfrom Iceland are biometrically more distinct fromflammea than exilipes is (Fig. 7).

On plumage, several pale birds from Icelandare similar to exilipes, some are even as bright ashornemanni, and the question arises if they do notmerely belong in one of these taxa. A canovar anal­ysis on all pale birds shows a chain of three popu­lations with highly different means (Fig. 8, K =5.8). Scores of separate DFAS are highly signifi­cantly different between pairs of samples with suf­ficient data (males): exilipes versus pale islandica(Mann-Whitney-U testZ= 5.9,P < 0.000001) andpale islandica versus hornemanni (Mann-Whit­ney-U test Z = 6.4, P < 0.00000 I). However, con-

1.3J--.-------r-----r---~--...,

Fig. 7. Plot of canonical variate analysis on wing, tail,bill-length and -depth ofj7.ammea (including "holboel­Iii", F, full line, N =438), exilipes (E, dashed, N =84)and pale islandica (P, dotted & dashed, N = 26). Samplemeans (+), centroids and 95% equiprobable ellipsesshown. Total eigenvalue: K =2.39.

'"p

0 1.6~

"~'":> 1.5 /~/--()

c.5 /, /, ,, \ /e I' II; G ,,;J J

1.4 I II I, ,, ,

, ,,

C.V. 1 (K= 2.27)

Taxonomic implicationsThe results concerning "holboellii" and the

sibling pairs flammea/exilipes and rostrata/hor­nemanni are in agreement with Molau's (1985) andKnox's (1988) recent conclusions. Knox alsoshowed these conclusions were not in conflict withearlier work on Redpolls, although some of the ear­lier data had to be reinterpreted. The overlap in bio­metry reported here betweenflammea and exilipes(ca.5%) is very similar to that reported by Troy(1985), based on skeletal measurements, betweenhis samples of "pure" flammea and exilipes.

Vaurie (1956, 1957) questioned the distinctionof islandica, after apparently only examining darkbirds. He noted a small, yet inconsistent differencefrom rostrata. For the occurrence in so small a pop­ulation as is only found in part oflceland of a varie­ty of plumage phenotypes the extremes of whichare elsewhere only seen among species three pos­sible conclusions may be suggested: (1) two spe-

0.90.80.70.60.5

TAXONOMY IN REDPOLLS 449

cies occur in Iceland; (2) the Iceland population re­presents a hybrid swann between dark rostrata andpale hornemanni (Salomonsen's (1951) sugges­tion, see also Wolters 1979); (3) the isolated popu­lation has been subject to considerable characterrelease resulting in unusual variation.

Let us first examine the last possibility. Basedon the significant difference in tail-length, Knox(1988) concluded that the dark and pale birds inIceland did not simply represent colour-morphs,and this also rules out character release. The sameconclusions follow from Fig. 6: average biometryis different between pale and dark birds. The link­age of plumage phenotypes (possibly under thecontrol of major genes) and differences in averagebiometry (likely controlled by a number of additiveeffects) in Icelandic Redpolls makes the possibilityof a hybrid swann less likely, unless incompleteintrogression and assortative mating occur. How­ever, in the White-throated Sparrow Zonotrichiaalbicollis, a species which genetics are far betterunderstood than those of Redpolls, the two plum­age morphs are documented to differ also in severalbody measurements (Rising & Schields 1980),among other characters. Comprehensive geneticresearch of Icelandic Redpolls will help to clarifythe genetic basis of this most confusing situationand it may reveal the rate of gene flow, if any. Thepresence of indetenninate birds in my classifica­tion mainly results from an oversimplification bydefining dark and pale birds as rigid categories. Ifsuch rigid colour descriptions are to be applied toplumages in different seasons and to birds of dif­ferent age categories, it is clear that many speci­mens do not match either dark or pale and will beclassified as indetenninate. The situation resem­bles the traditional thinking about fiammea andexilipes, which is now believed to be incorrect.Oversimplified definitions by typical plumagesand underestimation of the real intraspecific varia­bility, e.g. due to sex and age (Molau 1985), hadlead to suggestions ofhybridization. Fig. 5 ofKnoxapplies also to the Iceland situation. That I onlyfound 5 pale females in Iceland is reminiscent ofthe situation constructed by Troy (1985) for theNearctic, where he only found two females of exi-

lipes in a sample selected on paleness of the plum­age. Knox (1988) pointed out that the Troy-paradoxfonns a general model, explained by two separatepopulations with overlap of characters betweensome age and sex classes. It can be concluded thatthe Iceland population most likely consists of birdswhich genetically belong either to the pale or to thedark form. Knox's own observations that mostbirds could "easily be assigned to either the lightor the dark fonn" during the breeding season ismost significant (Knox 1988).

The population in Iceland seems to consist onthe one hand of birds very similar to rostrata (darkislandica) and on the other hand of exilipes-likebirds, ranging between rather heavily streaked phe­notypes (very fiammea-like; indetenninate here)and very pale phenotypes (pale islandica) muchlike "true" exilipes or hornemanni. Phenotypically,several museum specimens are impossible to allo­cate: the situation is especially complex due to theeffect of age and of the seasonal changes in plum­ages which are greater than the inter-taxa differ­ences. It can, however, not be excluded that somebirds are true hybrids. My interpretation is that inIceland an exilipes-like pale taxon has been subjectto character release in its evolution towards anendemic fonn before it came more recently intocontact with the second invader rostrata. Ifa hybridswann has then started to develop cannot be con­cluded at this stage of knowledge.

The question now arises to which bird the ori­ginal name islandica was given. Unfortunately, inhis original description of islandica, Hantzsch(1904) did not indicate a holotype, but used a sam­pIe of ten birds. I was able to relocate only 3 spe­cimens of Hantzsch's original series (the nOs VII,VIII and X in his publication, now in ZMHU,Berlin). Although Hantzsch stated that all ten con­stituted a homogeneous series, in my classificationVII and X are dark fonns, but, as far as can bejudged from its present state, VIII rather seems in­detenninate. According to Hantzsch VII and VIIIwere breeding mates. However, Hantzsch mostsurprisingly mentions a pale rump ("btirzel weiss­lich"), which hardly fits dark fonns, and he dis­tinguishes between paler and darker birds ("hel-

450 ARDEA 78 (3), 1990

leren" and "dunkleren Eksemplaren"). It is possi­ble and likely that his description applies to a he­terogeneous series. Consequently, the name islan­dica could be considered as indeterminate.However, to conserve stability in nomenclature,

I advocate the use of islandica Hantzsch for thedark rostrata-like Iceland form and I restrict its usethereto. In doing so it is desirable to chose a lecto­type and I designate his N° VII = n° 79.107 inZMHU, Berlin. This specimen was collected byHantzsch on 18 July 1903 north of Reykjalid, My­vatn (Hantzsch's spelling).

In Iceland there is therefore an apparently dis­continuous population of dark forms very near torostrata (true islandica) and paler forms most ofwhich are like dark individuals of exilipes, al­though some are as bright as hornemanni. Bio­metrically, the paler birds on Iceland can hardly bethe result of introgression following hybridizationbetween any dark and pale taxon (Fig. 1). Unlessstrong assumptions are made concerning foundereffects, the pale birds from Iceland constitute aseparate entity in the chain ofpale taxa, in betweenexilipes and hornemanni (Figs. 1 and 8). If there isa population of clearly distinct Redpolls in Iceland,it comprises the pale birds. Whatever the state ofhybridization or introgression may be, there is evi­dence for the existence of an endemic pale taxonin Iceland. If gene flow between the two popu­lations can be shown to be reduced, the pale birdsshould be given a new name.

On measurements and plumages the two pop­ulations in Iceland are very close to each other andare even closer than the dark and pale birds are inthe other two species pairs. Further research mayreveal criteria which could better separate them andsuggest if etho-ecological or other mechanisms ac­tually isolate them. It is possible that they aremicro-geographically isolated by habitat as in exi­lipes vs. flammea (Molau 1985, though see alsoNystrom & Nystrom 1987). In the few other casesin the world where two very similar taxa occur ona small island (or in restricted areas of largerislands), a (micro)-geographical segregation,enhanced by habitat and/or altitudinal preferencesis apparent. On two of the Comoro Islands there is

a pair of parapatric bulbul species Hypsipetessegregated by altitude and habitat (Louette &Herremans 1985). The two endemic Jamaican hum­mingbird species Trochilus are allopatric, but alsoshow differences in abundance according to alti­tude and habitat (Schuchmann 1978). Two speciesof brush warblers Nesillas on the small island ofMoheli have slightly different altitudinal prefer­ences, but mainly differ in niche (Louette et al.1988). The situation of the Mascarene White-eyesZosterops borbonica on Reunion Island seemsmore complex, but a segregation of phenotypesaccording to altitude and/or habitat is also apparent(Gill 1973).

Radiation and relationships among RedpollsGeneral The two roughly isometric groups in Fig.1, which correspond with the dualism in generalcolour appearances seem to support the taxonomicview of two polytypic species. However, similarityin general paleness is a poor criterion to use fortaxonomical conclusions in Redpolls. All Redpollsare so much alike in general colour pattern, thatany isolated populations evolving pale plumagewould become very similar. Paleness is a featurethat is expressed in all populations in a parallel way:it always consists ofmore whitish parts in the plum­age (especially rump, underside, nape and wingmarkings) and less heavy streaking, especially onthe paler parts. Even within taxa, pale individualscause confusion: pale specimens of cabaret verymuch resemble flammea (see e.g. the discussionabout the origin of the pale New Zealand cabaretspecimens: Stenhouse 1962 versus Fennell et al.1985), and pale flammea is very difficult to dis­tinguish from exilipes (see Troy 1985 versus Molau1985 and Knox 1988). Furthermore, only the paleand less streaked specimens of exilipes resemblehornemanni, while others are moreflammea-like.

The biometrical duality shown in Fig. 1 doesnot constitute a strong taxonomic basis either. Itcould merely represent ecological convergence:shorter-tailed, longer-billed birds that live in trees(cabaret andflammea) are in the right chain, andbirds that adapted to more open, windy countryshow longer tails and shorter bills (possibly also

TAXONOMY IN REDPOLLS 451

related to more ground feeding) and constitute theleft chain. The darkest Redpoll, rostrata, living inthe relatively open arctic, also tends towards theleft chain. In fact, within the subgenus Acanthis,the Twite Carduelis flavirostris, which is alsoadapted to open, nearly treeless country, has a veryshort bill and a very long tail, representing a similarconvergent evolution. The development of paleplumage in the open country taxa situated on theleft chain is likely to be a co-adaptation to arcticconditions, simply following Gloger 's rule, and notnecessarily indicating close genetic relation- ship.The adaptation of pale plumaged taxa to more arc­tic conditions is also shown by their convergentfl uffiness, resulting in better insulation and there­fore broader temperature tolerance (Brooks 1968).

On bill morphology, rostrata and hornemannifrom Greenland are distinct from all other Red­polls. Besides being the possible result of a split ofthe same old tribe in refugia in the extreme north­western Palearctic (e.g. Peary vs. coastal Green­land; Salomonsen 1970) this may also be the resultof an obligate convergent adaptation to meet spe­cial feeding conditions on Greenland, and so wouldmask earlier morphological differences.

No morphological criteria yet investigatedseem to be useful to build a strong taxonomy in theRedpolls.

Zoogeography Redpolls either radiated from acommon ancestor during only one step of geogra­phical isolation, or they are the result of a multiplestep radiation. If all recent Redpoll taxa are the re­sult of isolation during the last glacial period(Wiirm-Wisconsin) only, they must all be of aboutthe same age. Consequently, they could all haveacquired an equivalent taxonomic position: e.g. ifthe split between cabaret, flammea and exilipesdates from that period, cabaret may well be speci­fically distinct from those siblings. It also impliesthat the palest Redpoll (hornemanni) and the dark­est one (rostrata) should have radiated from a com­mon ancestor in only one step, both under arcticconditions. However, according to Marten & John­son (1986) the Wiirm- Wisconsin glacial periodwould be much too recent an event to account for

the genetic distance found between flammea andexilipes. Redpolls more likely radiated in more thanone step.

The actual distribution offlammea and exilipesappears to represent a rapid and successful spreadthrough the contemporary interglacial boreal-arc­tic zone. Their long-distance migratory move­ments often show an east-west component (Troy1983, Holgersen 1982, Runde 1984, 1985, 1987).This may represent a recapitulation of the coloni­zation direction. Although I have not studied thisaspect in detail, it seems as ifover their huge range,bothflammea and exilipes show no clear geogra­phical differentiation, pointing to the rather recenthistory of this distribution and/or to extreme gene­flow. Because of similarities in colour and biome­try and because these apparently show so little vari­ation over the huge breeding ranges,flammea andexilipes could be regarded as the most recent sib­ling pair of Redpolls, separated only from the lastglacial period onwards. However, some biochemi­cal evidence indicates that separation offlammeaand exilipes could be about 550 000 years old(Marten & Johnson 1986). Even if this date is onlyapproximately correct, the two taxamust have beenseparated during an earlier glaciation. It is, how­ever, very unlikely thatflammea and exilipes havebeen as successful during previous interglacialperiods as they are now, because if they have hadhuge ranges extending over the hoIarctic at suchperiod, they should at least have radiated into poly­typic species during the subsequent glaciation. Thetwo taxa breeding on Greenland, hornemanni androstrata, could indeed be considered as the resultof such, and consequently they may have evolvedfrom exilipes andflammea, respectively, during thelast glaciation. The distinctiveness of the Green­land birds from their ancestors could be the resultof enhanced evolution, not unlikely when verylimited populations have been subjected to strongadaptive selection under peculiar insular condi­tions. The resemblance in bill morphology betweenhornemanni and rostrata could be the result ofcon­vergent adaptive selection to meet special feedingconditions in Greenland. The striking differencesin colour and size could be interpreted as character

452 ARDEA 78 (3),1990

displacement, which in itself could be a proof ofspecific status for rostrata versus hornemanni.However, the colour resemblance of rostrata tocabaret is striking: both are very dark, heavilystreaked, the dark brownish hue is especially dis­tinctive. The occurrence of the much darker ro­strata more to the north of flammea is strange,because it is in disagreement with the ecogeogra­phical rule of Gloger, otherwise valid among Red­polls. It is possible thatflammea was already dif­ferentiated before the last glacial period in a dark,westerly proto-cabaret and a paler, easterly proto­flammea. rostrata could have originated from thedark proto- cabaret, occurring in the western pale­arctic. The overall parallel in biometrical evolutionis remarkable if one accepts that hornemannievolved from a proto-flammea/exilipes ancestor,and rostrata from proto- cabaret (see Fig. 1). Nocharacter displacement would then be necessary toaccount for the present phenotypes. In view of therapid and for Redpolls rather dramatic changes thathornemanni and rostrata went through, they canbe argued to have reached specific status.

islandica is possibly merely a synonym ofrostrata and the tendency for a smaller bill andpaler plumage can be seen as the result of (presentor ancient) partial introgression from pale birds. Ifno introgression was involved, islandica would bea poorly differentiated population of rostrata thatbecame settled on Iceland only after the last gla­ciation. Pale birds from Iceland show affinities toexilipes and hornemanni, but seem closer to exi­lipes, from which taxon they may have evolvedduring the last glaciation. Their more striped plum­age could be interpreted as an introgression fromrostrata/islandica, but biometrically they are ob­viously in the chain of pale taxa. Intuitively onemay expect pale birds from Iceland to be moreclosely related to geographically close horneman­ni from Greenland. However, the zoogeographicalpicture in other landbirds in the northwesternPalearctic is not very consistent and helpful in cla­rifying the most likely affinities of the pale birdsin Iceland. In the Lappland bunting Calcarius lap­ponicus European and Greenland birds are in thesame subspecies, and the same holds for the Snow

bunting Plectrophenax nivalis, but here Iceland hasan endemic subspecies. The resident PtarmiganLagopus mutus has been split in several more sub­species:one each in Spitsbergen, Iceland, Scandinavia,eastern, western and northwestern Greenland,among other regions. A zoogeographical similarityis found between the Redpolls and the Bean/pink­footed Goose complex: in the geese, there is a dis­tinct species Anser brachyrhynchus in the extremenorthwestern palearctic (Greenland, Iceland,Spitsbergen) and there are two paraspecies in Scan­dinavia and adjacent USSR (jabalis/rossicus), butfurther differentiation also occurred farther to theeast. Differentiation within exilipes or flammea ispossibly still to be discovered, and the affinitiesshould be studied also in the other pale Redpollsapparently existing on islands in the extreme north­western Palearctic (e.g. Spitsbergen, Jan Mayen).The existence there of hornemanni-like birds hasbeen mentioned repeatedly (Fischer & von Pelzeln1886, Ridgway 1901, Hartert 1903, Nathorst 1915,Van Franeker et al. 1986). It cannot be ruled out thatsome may constitute permanent populations andcould be endemic offshoots close to exilipes orhornemanni.

The present distribution of cabaret in Europeis that of a postglacial alpine relict, which probablyonly recently colonized England and some of thenearby coastal areas. In my opinion its habit ofreturning to the continent on migration seems arecapitulation of the colonization route from theglacial refugium in Europe, rather than the oppo­site. Knox (1988) and Ernst (1988), on the contrary,argue that the Alps might have been colonized fromBritain. If cabaret is older than the last glaciation,as suggested above to account for its apparentaffinities to rostrata, a careful study of all isolated(alpine) populations of cabaret possibly may showdifferences among them. The relationship of ca­baret to flammea is close, but unclear. The recentexpansion of the former in southern Scandinaviamay lead to contact with the latter. In view of dif­ferences in some of the vocalizations (Herremans1989), the social interaction of these populationsmay be limited, and probably also the gene flow in

TAXONOMY IN REDPOLLS 453

this future area of sympatry. Such could supportthe view that cabaret split off from flammea al­ready before the last glaciation.

Although there is as yet no objective basis tomeasure Redpoll affinities, the following taxono­mic arrangement of the Redpolls is proposed,mainly in the light of zoogeographical argument:

Carduelis (h.) hornemanniCarduelis (h.) exilipesCarduelis (h.) ssp. (pale Icelandic)Carduelis (f) rostrata (inc!. islandica)Carduelis flammea flammeaCarduelis flammea cabaret

Because most morphological criteria seem to con­stitute only poor evidence of the affinities in thiscomplex group, field studies of etho- eco-pheno­logical segregation may be of more interest in fu­ture. Detailed studies by biochemical techniquesare also urgently needed: comprehensive knowl­edge ofgenetic distances seem ofparticular interestto quantify the affinities within this group, whichapparently is phenotypically very conservative.

ACKNOWLEDGEMENTS

I am most grateful for hospitality in the Koninklijk Museumvoor Midden Afrika, Tervuren (Belgium): Dr. M.Louettearranged loans of specimens and Dr. D.Meirte was mostcooperative with multivariate analyses using his personalroutines. The staff of the BMNH, the ZMK and the KBINallowed access to the collections; especially Dr. A.G.Knox,Dr. J.Fjeldsa and w'Roggeman were very helpful and assis­ted in various ways to make these visits pleasant and suc­cessful. lowe many thanks to the curators offollowingmuse­ums for shipment of specimens: AY.Andors (AMNH),B.McGowan (RMS), KSkipnes (SMN), G.Mauersberger(ZMHU), and A.Petersen (IMNH). Several more museumsare thanked for informative correspondence concerningtheir collections. The general concepts about Redpolls havegreatly benefited from discussions and correspondence withA.G.Knox, who is particularly acknowledged for his gene­rous cooperation. The manuscript has been improved byM.Louette, D.Meirte, and very much so by A.G. Knox. Prof.KH. Voous and an anonymous referee are acknowledgedfor valuable discussion.

REFERENCES

Bird, E.G. 1935. The status of Redpolls in Iceland. IbisV (l3th Series): 438-441.

Bird, E.G. 1936. Further note on Redpolls in Iceland.Ibis VI (l3 th Series):381.

Blackith, R.E. & R.A Reyment 1971. Multivariate mor­phometries. Academic Press, London.

Brooks, W.S. 1968. Comparative adaptations of theAlaskan Redpolls to the arctic environment. WilsonBull. 80: 253-280.

Ernst, S. 1988. Die Ausbreitung des A1penbirkenzeisigs,Carduelisflammea cabaret P.L.S. MUller, in Europabis zum Jahre 1986. Mitt. Zool. Mus. Berl. 64(Suppl.): Ann. Om. 12: 3-50.

Fennell, J.EM., P.M. Sagar & J.S. Fennell 1985. Varia­tion within the Redpolls ofCanterbury. Notornis 32:245-253.

Fischer, E & A. vonPelzeln. 1886. Vogel von Jan Mayengesamme1t von Dr. E Fischer. Mitt. Ornith. Ver.Wien 10: 194-197.

Gill, EB. 1973. Intra-island variation in the MascareneWhite-eye Zosterops borbonica. Am. Ornith.Union, Ornith. Monogr. n° 12.

Greenstadt, J. 1967. The determination of the characte­ristic roots of a matrix by the Jacobi method. In: A.Ralston & M.S. Wi1f (eds.) Mathematical methodsfor digital computers. John Wiley & Sons, London.

Hantzsch, B. 1904. Acanthis linaria islandica subsp. nov.Ornith. Monatsber. 12: 32-34.

Hartert, E. 1903. Die Vogel der paHiarktischen Fauna.Heft 1. Friedlander und Sohn, Berlin.

Hebrant, E 1974. Problemes de discrimination dans Iecas de plusieurs populations. Biometrie-Praxime­trie XIV: 15-41.

Herremans, M. 1985. Post-mortem changes in morpho­logy and its relevance to biometrica1 studies. Bull.Brit. Om. Cl. 105: 89- 91.

Herremans, M. 1987. Het voorkomen van "langsnave1­types" holboellii in de invasie Grote BarmsijzenCarduelisfflammea in 1986/87. Orio1us 53: 149­153.

Herremans, M. 1989. Vocalizations of Common, Lesserand Arctic Redpolls. Dutch Birding 11: 9-15.

Ho1gersen, H. 1982. Bird-ringing report 1979-80. Sta­vanger Museum. Sterna 17 (4): 85-123.

Knox, AG. 1988. The taxonomy ofRedpolls. Ardea 76:1-26.

Louette, M. & M. Herremans 1985. Taxonomy and evol­ution in the bu1bu1s (Hypsipetes) on the Comoro Is­lands. Proc. Int. Symp. Afriean Vertebrates, system­atics, phylogeny and evolutionary ecology (Bonn):407-423.

Louette M., M. Herremans, L. Bijnens & L. Janssens1988. Taxonomy and evolution in the Brush

454 ARDEA 78 (3), 1990

warblers Nesillas on the Comoro Islands. Tauraco1: 110-129.

Marten, J.A. & N.K Johnson 1986. Genetic relation­ships of North American Cardue1ine finches. Con­dor 88: 409-420.

Mayr, E. 1963. Animal species and evolution. HarvardUniversity Press, Cambridge.

Mayr, E. & L.L. Short 1970. Species taxa of NorthAmerican birds. Publ. Nutall om. Cl. No 9.

Molau, U. 1985. Grasiskkomplexet i Sverige. Var Fagel­varld 44: 5-20.

Nathorst, A.G. 1915. Den hognordiska grasiskans(Acanthis hornemanni Holb.) forekomst pa Spetz­bergen. Fauna och Flora 10: 204-207.

Nystrom, B. & H. Nystrom 1987. Biotopva1 och hack­ning hos grasiskorCarduelisflammea och snosiskorC. hornemanni i Ammarnasomradet sodra Lapp­land. Var Fagelvarld 46: 119-128.

Ridgway, R. 1901. The birds ofNorth and Middle Ameri­ca. Vol I. Bull. U.S. Nat. Mus. n° 50(1), Washington.

Runde, O.J. 1984. Bird-ringing report 1981. StavangerMuseum. Sterna 17 (5): 129-155.

Runde, OJ. 1985. Bird-ringing report 1982. StavangerMuseum. Sterna 17 (6): 157-185.

Runde, O.J. 1987. Bird-ringing report 1983. StavangerMuseum. Sterna 17 (7): 197-224.

Salomonsen, F. 1951. The birds of Greenland. Part III.Munksgaard, Kjjbenhavn.

Salomonsen, F. 1972. Zoogeographical and ecologicalproblems in arctic birds. Proc XV Intern. Ornith.Congr. (Hague): 25-77.

Schuchmann, K-L. 1978. Allopatrische artbildung beider Kolibrigattung Trochilus. Ardea 66: 156-172.

Sokal, R.R. & F.J. Rohlf 1969. Biometry (second edi­tion). Freeman and Company, New York.

Stenhouse, D. 1962. Taxonomic status of the New Zea­land Redpoll Carduelis flammea: a reassessment.Notornis 10: 61-67.

Svensson, L. 1984. Identification guide to European Pas­serines. (Third, revised and enlarged edition).Published by the author, Stockholm.

Todd, W.E.c. 1963. Birds of the Labrador Peninsula andadjacent areas. Toronto Press, Toronto.

Troy, D.M. 1983. Recaptures of Redpolls: movementsofan irruptive species. J. Field. Ornith. 54: 146-151.

Troy, D.M. 1985. A phenetic analysis of the RedpollsCarduelis flammea flammea and C. hornemanniexilipes. Auk 102: 82-96.

Van Franeker, J.A., K Camphuysen & F. Mehlum 1986.Status over Jan Mayens fugler. Var Fuglefauna 9:145-158.

Vaurie, C. 1956. Systematic notes on Palearctic birds.N° 19. Fringillidae: the genera Fringilla, Serinus,Carduelis and Acanthis. Am. Mus. Nov., N° 1775:1-25.

Vaurie, C. 1957. On the validity of Acanthis flammeaislandica Hantzsch. Dansk. Ornith. For. Tidsskr. 51:9-11.

Voous, KH. 1977. List of recent holarctic bird species.Ibis 119: 376-406.

Wolters, H.E. 1979. Die Vogelarten der Erde. 4 Liefe­rung. Paul Parey, Hamburg/Berlin.

SAMENVATTING

Barmsijzen werden taxonomisch onderzocht door devleugellengte, staartlengte, snavellengte en snavelhoogtevan meer dan 1000 gesexte individuen nauwkeurig temeten. Alle erkende taxa waren vertegenwoordigd. Demeetresultaten werden met technieken uit de multivariatestatistiek (Principale componenten analyse enCanonischeanalyse) verwerkt. Grote Barmsijzen (jlammea) en KleineWitstuitbarmsijzen (exilipes) die fenotypisch op kleurwaren gedetermineerd konden ook biometrisch geschei­den worden; hetzelfde gold voor de 2 Groenlandse taxa(rostrata en hornemanm). Vogels die voldeden aan de cri­teria van Langsnavelbarmsijzen ("holboellii") waren zogoed als identiek met flammea: zij vertegenwoordigenblijkbaar slechts een extreme scheefheid van de snavel­lengte-frequentieverdeling bijflammea. Het donkere typebarmsijs van Usland is nauwelijks te onderscheiden vande op Groenland voorkomende vorm rostrata, maargemiddeld zijn ze zowel biometrisch als wat het veren­kleed betreft verschillend van het bleke type dat op Uslandbroedt. Bovendien zijn de bleke barmsijzen van Uslandbiometrisch verschillend van de twee andere, allopatri­sche taxa met bleke barmsijzen (exilipes en hornemanni),maar ze staan het dichts bij exilipes. Bij de originelebeschrijving van islandica werd er geen holotype aange­duid en de gebruikte serie vogels was mogelijk hete­rogeen. Omdat wellicht de donkere vorm bedoeld werd,wensen we voortaan de naam islandica tot deze vogels tebeperken. Ze lijken sterk op rostrata; een lectotype voorislandica werd aangewezen. Van de beide vormen die opUsland voorkomen zijn het de bleke vogels die het meestvan alle andere barmsijzen verschillen: mogelijks verdie­nen ze een apparte, nieuwe naam.

De gebruikelijke taxonomische opsplitsing van debarmsijzen in twee polytypische soorten heeft een labieleonderbouw, omdat het verschil of de overeenkomst inmorfologische criteria het gevolg kan zijn van convergen­te ontwikkelingen in de evolutie. Vooral op basis van zoo­geografische argumenten kan de status soort worden toe­gekend aan hornemanni, exilipes, rostrata enflammea.

APPENDIX

TAXONOMY IN REDPOLLS 455

Sexingtaxon cabaret: In this taxon adults showing no pink(except at poll) or at most some pinkish spottings in thecheeks and eventually also vaguely on the breast wereconsidered females. Injuveniles, females included birdswithout pink and with at most some brownish-pink spotsin the cheeks. All birds with extensive pink on the cheeksand breast were considered males. Juveniles with pinkcheeks and clearly pinkish spots on the breast were alsoascertained as males.

Descriptive statistics:

taxon flammea: Females of this taxon have much morefrequently and extensive pink in the plumage: only birdswith extensive pink on cheeks and underside were con­sidered males. Females include juveniles without pinkon the underside, but also with at most some pinkishspots in the cheeks, and adults without pink or with atmost some pink spots on cheeks, rump or underside.Several birds did not fit in these categories and were notsexed.

Weight

Taxon Sex N Mean SD CV(%) Range

flammea M 181 13.3 0.96 7.2 11.0-16.0F 141 12.5 0.80 6.4 10.3-14.5

"holboellii"-form M 18 14.9 1.05 7.1 13.0-16.8F 15 13.5 0.88 6.5 12.0-15.3

Wing length

Taxon Sex N Mean SD CV(%) Range

hornemanni M 47 85.5 2.43 2.8 81.0-92.0F 42 82.7 2.13 2.6 79.5-88.5

pale islandica M 21 79.2 1.90 2.4 76.5-84.5F 5 77.4 1.82 2.4 75.0-80.0

indeterminate islandica M 40 78.2 1.93 2.5 72.5-82.5F 23 77.5 2.44 3.2 74.0-81.5

dark islandica "breeding" M 14 78.6 2.54 3.2 75.0-83.0F 11 75.5 2.19 2.9 72.0-78.5

dark islandica "not breeding" M 22 80.1 1.97 2.5 77.0-84.0F 20 77.1 2.08 2.7 73.5-81.0

rostrata "breeding" M 58 80.3 1.96 2.4 74.0-86.0F 25 78.8 2.65 3.4 75.0-85.0

rostrata "not breeding" M 43 80.3 2.00 2.5 75.5-86.0F 45 78.7 2.12 2.7 73.5-83.0

exilipes M 47 76.3 1.72 2.3 72.0-81.0F 37 73.6 1.99 2.7 70.5-78.0

cabaret M 66 71.1 1.92 3.1 67.0-75.0F 34 69.4 1.58 2.3 67.0-74.5

flammea M 222 77.6 2.18 2.8 70.5-82.5F 162 75.2 1.75 2.3 70.5-80.0

"holboellii"-form M 31 79.2 2.55 3.2 74.0-83.5F 23 76.5 1.03 1.4 75.0-78.5

456 ARDEA 78 (3),1990

Tail length

Taxon Sex N Mean SD CV(%) Range

hornemanni M 47 65.5 2.44 3.7 59.5-70.0F 42 64.4 2.69 4.2 58.0-69.5

pale islandica M 21 61.5 2.15 3.5 57.5-66.5F 5 61.0 2.70 4.4 58.5-65.0

indetenninate islandica M 40 60.2 2.53 4.2 54.5-65.0F 23 60.8 2.72 4.5 54.0-65.5

dark islandica "breeding" M 14 58.4 2.68 4.6 55.5-63.5F 11 57.6 1.98 3.4 54.5-60.5

dark islandica "not breeding" M 22 60.4 2.19 3.6 57.5-65.5F 20 58.5 2.44 4.2 54.5-63.0

rostrata "breeding" M 58 59.2 2.74 4.6 53.0-66.5F 25 59.0 2.75 4.7 54.5-66.0

rostrata "not breeding" M 43 59.5 2.83 4.7 52.5-65.0F 45 59.2 2.71 4.6 52.0-64.5

exilipes M 47 58.1 1.67 2.9 54.0-61.5F 37 57.4 1.60 2.8 53.5-60.5

cabaret M 66 52.3 1.59 3.0 48.0-56.0F 34 51.5 1.56 3.0 47.5-54.5

flammea M 222 55.2 2.04 3.7 49.0-60.5F 162 54.2 1.89 3.5 49.0-59.5

"holboellii"-fonn M 31 55.4 1.86 3.4 53.5-61.0F 23 56.7 1.40 2.5 54.0-60.0

Bill-length

Taxon Sex N Mean SD CV(%) Range

hornemanni M 47 8.94 0.58 6.5 7.8-10.6F 42 8.67 0.52 6.1 7.7-10.0

pale islandica M 21 8.24 0.37 4.5 7.7- 9.3F 5 7.92 0.13 1.7 7.7- 8.0

indetenninate islandica M 40 8.51 0.43 5.1 7.4- 9.3F 23 8.40 0.65 7.8 7.9- 9.1

dark islandica "breeding" M 14 9.13 0.54 5.9 8.2- 9.9F 11 8.78 0.50 5.7 8.0- 9.4

dark islandica "not breeding" M 22 8.73 0.48 5.6 8.0- 9.6F 20 8.66 0.67 7.7 6.6-10.0

rostrata "breeding" M 58 9.47 0.55 5.8 8.3-11.0F 25 9.32 0.57 6.1 7.8-10.2

rostrata "not breeding" M 43 9.11 0.49 5.4 8.0-10.0F 45 9.17 0.37 4.0 8.4- 9.9

exilipes M 47 7.87 0.45 5.8 6.8- 8.8F 37 7.71 0.47 6.1 6.3- 8.5

cabaret M 66 9.01 0.46 5.2 7.7-10.2F 34 8.81 0.36 4.1 8.1- 9.5

flammea M 222 9.13 0.59 6.5 7.9-10.4F 162 8.83 0.56 6.3 7.5- 9.9

"holboellii"-fonn M 31 10.67 0.45 4.2 10.5-11.6F 23 11.10 0.55 5.0 10.0-12.9

TAXONOMY IN REDPOLLS 457

Bill depth

Taxon Sex N Mean SD CV(%) Range

hornemanni M 47 7.11 0.24 3.4 6.3-7.5F 42 6.88 0.28 4.1 6.1-7.8

pale islandica M 21 6.32 0.20 3.2 6.0-6.8F 5 6.18 0.08 1.4 6.1-6.3

indetenninate islandica M 40 6.43 0.27 4.2 5.9-7.0F 23 6.33 0.06 4.2 5.7-6.8

dark islandica "breeding" M 14 6.61 0.22 3.4 6.2-7.1F 11 6.33 0.19 2.9 6.0-6.6

dark islandica "not breeding" M 22 6.65 0.28 4.2 5.9-7.1F 20 6.46 0.34 5.3 5.7-7.0

rostrata "breeding" M 58 6.68 0.26 3.9 6.3-7.5F 25 6.64 0.25 3.7 6.3-7.3

rostrata "not breeding" M 43 6.59 0.21 3.2 6.2-7.0F 45 6.46 0.23 3.6 5.9-7.2

exilipes M 47 6.03 0.19 3.2 5.6-6.4F 37 5.92 0.22 3.8 5.3-6.3

caharet M 66 5.99 0.23 3.8 5.5-6.5F 34 5.82 0.26 4.5 5.2-6.5

flammea M 222 5.91 0.25 4.3 5.3-6.8F 162 5.73 0.24 4.2 5.2-6.5

"holhoellii"-fonn M 31 6.28 0.19 3.0 5.8-6.8F 23 6.45 0.29 4.5 6.1-7.2

Canovar transformations for selected taxa

The correct use of these functions and the cor­responding figures in the identification of individualbirds is highly dependent on the measuring procedures.Please consult the method section and try to obtain be­forehand similar average measurements for at least onetaxon as detailed as in tables given here (alternatively, acorrection factor should be applied to the data).

X =maximum wing-length, Y =tail-length, Z =bill­length from feathering, V = bill-depth at feathering.

The coefficient of determination (% in brackets)indicates how the distinctive power of each variable isdivided between the canonical axes.

A. Canovar onflammea and exilipes (Fig. 2):CV 1 = -0.006 X (22%) + 0.012 Y (81.6%) -0.089 Z(94.8%) + 0.1196 V (27.1 %)CV 2 = 0.020 X (78%) -0.001 Y (18.2%) -0.018 Z(4.8%) + 0.106 V (71.2%)

B. Canovar on rostrata and hornemanni (Fig.4):CV 1 = 0.010 X (95%) + 0.0099 Y (99%) -0.0652 Z(69.2%) + 0.158 V (96.8%)CV 2 = 0.026 X (5%) -0.0156 Y (1 %) + 0.0633 Z(29.7%)-0.0287 V (3.1 %)

C. Canovar on flammea, exilipes and pale islandica(Fig.7):

CV 1=-0.00436X(5%) +0.0136 Y(91.5%)-0.0632Z (92.4%) + 0.126 V (69.7%)CV 2 = 0.013 X (95%) + 0.0067 Y (8.5%) + 0.0146Z (7.6%)-0.010 V (30.3%)

D. Canovar on the pale taxa: exilipes, pale islandicaand hornemanni (Fig. 8):

CV 1 = 0.0055 X (99.9%) + 0.0152 Y (99.2%) ­0.0013 Z (100%) + 0.2014 V (99.7%)CV 2 = -0.0049 X (0.1 %)-0.0228 Y (0.8%) -0.0505Z (0%) + 0.253 V (0.3%)

458 ARDEA 78 (3),1990

E. Canovar on the dark taxa: cabaret, flammea and ~

rostrata including dark islandica (additional Fig. 9):I"-

0 F

CV I =0.0022X (51.7%) + 0.0068 Y(93.3%) -0.039 II 0.8 , ,ll::

..... ,... \ ,

2 (33.4%) + 0.1184 V (98.3%) ... ,/ \

CV 2 = 0.0145X (48.3%) +0.0001 Y (6.7%) + 0.012 N e I II 0 I

2(66.3%)-0.0834 V (1.7%) :> I

U 0.7 II

, I.-8 .-

R

0.6

Fig. 9. Plot of canonical variate analysis on wing, tail, cbill- length and -depth of cabaret (C, full line, N = 100),flammea (F, dotted & dashed, N = 438) and rostrata +dark islandica (R, dashed, N = 238). Sample means (+), 0.5

centroids and 95% equiprobable population ellipses0.8 0.9 1.0 1.1

shown. Total eigenvalue: K = 3.75. C.V.1 (K=3.01)